The Einstein-Podolsky-Rosen nonlocality puzzle has been recognized as one of the most important unresolved issues in the foundational aspects of quantum mechanics. We show that the problem is resolved if the quantum correlations are calculated directly from local quantities which preserve the phase information in the quantum system. We assume strict locality for the probability amplitudes instead of local realism for the outcomes, and calculate an amplitude correlation function.Then the experimentally observed correlation of outcomes is calculated from the square of the amplitude correlation function. Locality of amplitudes implies that the measurement on one particle does not collapse the companion particle to a definite state. Apart from resolving the EPR puzzle, this approach shows that the physical interpretation of apparently `nonlocal' effects like quantum teleportation and entanglement swapping are different from what is usually assumed. Bell type measurements do not change distant states. Yet the correlations are correctly reproduced, when measured, if complex probability amplitudes are treated as the basic local quantities. As examples we discuss the quantum correlations of two-particle maximally entangled states and the three-particle GHZ entangled state.